Life Prediction Under Biaxial Loading Based On The Growth Behaviour of Short Cracks
Life Prediction Under Biaxial Loading Based On The Growth Behaviour of Short Cracks
Life Prediction Under Biaxial Loading Based On The Growth Behaviour of Short Cracks
Abstract
Biaxial fatigue tests were performed with thick walled tubular specimens made out
of a ferritic structural steel and an austenitic stainless steel. Cyclic hardening/sof-
tening, crack initiation, crack growth, and cycles to failure were measured. The
cyclic stress strain behaviour was found to be close to the Masing behaviour. Thus
the AJ and the related damage parameter, Zip, were used to describe the growth of
intermediate cracks and the resulting cycles to failure. It was found that by using
these parameters the data obtained in tension, torsion and biaxial tests can be
described by one universal law.
1 Introduction
Life prediction for cyclic loading conditions has been very successful in laboratory
tests using uniaxially loaded laboratory specimens. But the ultimate goal of fatigue
research will always be the life prediction for real parts. They are, however, typi-
cally loaded under multi-axial conditions, especially at notches and other highly
loaded areas. One simple possibility to transfer the results obtained in uni-axial
loading to multi-axial loading is to replace the tensile components of stress and
strain by the equivalent stresses and strains according to von Mises. Following
Manson/Coffin the plastic equivalent strain would then be the essential parameter
determining the cycles to failure, Nf. However, with such a simple approach many
experimental findings cannot be described. Therefore, other parameters have been
VoL3,Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
proposed to improve the situation. Brown and Miller [1] proposed a generalized
theory based on the maximum shear strain and on the tensile strain normal to the
plane of maximum shear. A modification of this approach was proposed by Lohr
and Ellison [2] by claiming that the strain which is normal to the specimen surface
is of special importance. Kandil et al. [3] and Socie et al. [4] refined this approach
in order to improve the description of mean stress effects.
A criterion which relys on the applied stresses was developed by Krempl [5]. Gar-
rud [6] and Ellyin [7] proposed an energy criterium to determine the life under
biaxial loading. The latter approach is closely related to the use of the cyclic J-inte-
gral, Δ J, introduced by Dowling and Begley [8,9]. Wüthrich [10] proved that the Δ J
is indeed a path independent integral around the crack tip if the material shows
Masing behaviour. Heitmann [11] used the ΔΙ to derive a damage parameter suit-
able for life prediction for random uni-axial loading. The latter treatment seemed
so promising that it will be tried in this paper to extend the AJ approach to propor-
tional biaxial loading.
2 Experimental procedure
The experiments were done with an austenitic stainless steel, German grade
X6CrNil811 and with a ferritic construction steel, German grade StE 70. The
chemical composition is given in table 1. Tubular specimens with an O D of 18 mm,
a wall thickness of 4 mm, and a gauge length of 36 mm were used. Due to this large
wall thickness elliptical surface cracks with a constant aspect ratio of 0.9 could be
obtained up to a depth of 1 mm. The specimen geometry, which agrees with ASTM
standards, is shown in figure 1. The specimens from the austenitic stainless steel
were annealed in vaccum p < 1 0 " 4 m b a r and furnace cooled. The annealing para-
meters and the resulting grain sizes are shown in table 2. Twin boundaries were
ignored for the determination of the grain size. The grain size of the ferritic steel
was 13 Mm, that of the austenite 50 to 450 μηι, depending on the heat treatment.
C Cr Ni Si Mn Ρ s Mo Zr
X6 CrNi IS 11 0.05 17.5 10.C 0.45 1.6 0.02 0.002 0.02 —
104
J. Prochotta and P. Neumann Journal of the Mechanical Behavior ofMaterials
The surface finish of the specimens was a mechanical polish with 1 μιη diamond
paste. Additionally, the austenitic steel was electro-polished thereafter.
Since crack initiation can be observed only at the outer surface of the tubular
specimen, crack initiation at the inner surface was avoided by reducing the surface
roughness and introducing compressive stresses by honing. The tests were per-
formed in a servo-hydraulic closed-loop tension-torsion system for a maximum
load of 400 kN and a maximum moment of 4 kNm. The axial and torsional strains
were measured with the help of strain gauges which were attached to the inner sur-
face of the tubular specimens.
The control of the system and the recording of the data was performed by a P D P
11/73 computer from Digital Equipment. Stresses and strains were recorded with a
resolution of 12 bit. The experiments were carried out with an instantaneous strain
or stress control (reaction time 5 ms). Superimposed was an amplitude control with
a reaction time of a few cycle times. In this way stress, total strain or plastic strain
amplitude could be controlled independently for tension and torsion. Furthermore,
the compliances for tension and torsion could be determined with an accuracy of
14 bit by compensating for the elastic slope of the hysteresis.
r 54
CO I
/
Q.
J
Ο LO
— CM
36
200
54
36
200
105
Vol 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
A special problem is the undesired coupling of tension and torsion signals in stress
as well as in strain. Since the coupling was almost linear and constant in the used
ranges for the given load cell and a given specimen with its strain gauges, it could
be compensated by a simple analog circuit, which introduced an appropriate small
adjustable tension signal into the torsion signal and vice versa. If the actual forces
become larger than 85 kN, a quadratic coupling term becomes noticeable but
could still be neglected. Similarly, a quadratic coupling term became noticeable
with the ferritic steel at small plastic strain amplitudes ( Δ e p < 10 ~ 4 ) .It was also
negligible for strain resolutions down to 3 · 1 CT 6 .
Crack growth was measured in experiments with constant plastic strain amplitude
and constant mean stress. For the determination of cyclic stress strain curves,
incremental-step tests were performed during proportional biaxial tension-torsion
loading. The number of steps varied between 40 and 80. Their height was not con-
stant but was increased by a constant factor. In this way also in experiments with a
large maximum amplitude sufficientely small plastic strains could be reached with
the number of steps given above and in a figure with log scales the distances be-
tween the measured points are constant.
A few tension tests were carried out with solid cylindrical specimens for compari-
son with the tubular specimens, and no significant differences were found. In all
biaxial tests the proportional loading was done with a constant plastic strain biax-
iality. All tests were perfomed with a sinusoidal loading. The frequency was always
chosen in such a way to avoid excessive heating of the specimen. In this way the
frequency used ranged from 2 Hz down to 0.3 Hz. However, in tests with a large
torsional amplitude, e.g. 7t = 0.01, the frequency had to be reduced down to 0.01
Hz.
The crack growth was measured by taking replicas in regular intervals, typically 5%
of the expected life time (de Lange [12] and Heroux [13]). With carbon shadowed
Table 2: Heat treatment and resulting grain size of the steel X6 CrNi 18 11
106
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
acetate replica cracks longer than ΙΟμτη could be detected in the optical micro-
scope. From the surface length of the cracks the crack depth could be determined
by measuring their aspect ratio on the fracture surface using fracture markings. By
tracing back large cracks on the preceding replicas it was possible to determine the
crack growth history back to the crack initiation.
The end of the life was determined by the appearance of cracks which are longer
than 1 mm. In tension such cracks do significantly change the compliance which
may be used to stop the test automatically. In torsion, however, surface crack
lengths of more than 6 mm sometimes could not be detected by a change in the
compliance.
3 Results
Since the strains were measured by strain gauges glued onto the specimen, the
elastic moduli could be determined from the slope of the elastic range. Table 3
gives the results.
In tension-torsion tests the stresses are not constant over the cross section of the
specimen. Below yielding the torsional components are proportional to the radius.
When plasticity commences at the outer surface, the radial distribution of torsional
stresses becomes non-linear. Nadai has given a quantitative treatment of this prob-
lem for solid cylindrical specimens [14]. A solution for thick-walled cylindrical
tubes was deduced by extending Nadai's solution by Neumann and Prochotta [15].
Let Τ(γ 0 ) be the torque measured at the specimen as a function of the shear strain
at the outer surface, γ 0 . y0 is connected to the twist angle, Θ, the radius r, and the
length of the gauge section, 1, by
StE 70 X6 CrNi 18 11
E-Modul / GPa 215 204
G-Modul / GPa 84 80
Poisson's ratio 0.28 0.27
107
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
W(Y) = — ( 2 )
2nr
For tubular specimens with a ratio, c, between inner, rj, and outer, r 0 , radius
Γι
c=- (3)
'0
a solution according to [15] is given by
r(7o) = Σ + ^ T ^ T (4)
(4) is exact if η is taken large enough so that c " γ is in the elastic part of the torque-
twist curve. For the specimens used here with c = 5/9, η = 2 is sufficient for all
cases. (4) was used to calculate shear stresses from the torque-twist curve.
Figures 2 and 3 show the cyclic softening, i.e. a decrease of the torque amplitude,
in ferritic steel experiments with constant plastic shear amplitude and zero mean
οο
CO
οο
to
οο
τ
'<Ε
ο
\ ο
η
CM
Ρτ
CS
ο
ο
Figure 2: Amplitude of the torque as a function of the cycles during a torsion ex-
periment with controlled plastic shear strain amplitude for the steel StE-70.
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
torque. After less than 10% of Nf the cyclic softening is complete and the torque
amplitude saturates.
Cyclic stress strain curves, CSSC, are difficult to determine in thick-walled tubular
specimens because of the radial dependence of stresses and strains. In order to
avoid these difficulties some specimens were produced with an inner diameter of
16.5 mm resulting in a wall thickness of 0.75 mm. Such specimens are thin-walled
tubular specimens according to the ASTM standards. The differences between
elastic and fully plastic calculations are less than 2% of the radial stress distribu-
tions. CSSCs were determined for both steels in multi-level tests with increasing
amplitude under pure tension, pure torsion, and proportional tension-torsion load-
o
109
Vol 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
The data for the austenite show no dependence on the biaxiality if the equivalent
stresses are plotted vs. plastic equivalent strains. In the case of the ferritic steels
deviations of this type occur only at plastic strains below 0.03%. Therefore a uni-
fied description of the biaxial CSSC in terms of the equivalent stresses and strains
seems to be justified. The shape of the CSSC shows two regimes with different
hardening exponents. Therefore, they may be described by a combination of two
power laws according to
The constants which are obtained from figures 4 and 5 are given in table 4.
ο
5 ! 1 1 1 1 1 III 1 1 1 1 1 1 III 1 1 I I IIII
X6 CrNi 18 11
ο
3 S
/A
• Tension
< Ο C Torsion
Κ Δ biaxial,Τλ..ρ = v'3
R = -1
I 1 I Μ 1 III 1 1 1 1 1 1 Ml ι ι I II 1 II
1 3
10- lO"· ΙΟ"
ΔΖρ/2
Figure 4: Cyclic stress strain curves according to multi level tests of the steel
X6 CrNi 18 11.
110
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
o
ΙΟ
CO 1 1 1 1 1 1 III 1 1 I I 1 lll| 1 ι I 1 1 1II,
StE-70 J
/
ο A/
ο / -
CO
/
A/°
β /
α, ο • tension ο /
w, ΙΟ A
r
ΙΟ
Ο torsion /a
Δ biaxial, \ f p = x/3 y
CM
ο A^
lb ο
<1 a *
ο - *
ο
I - " ° li = -1
ο " ο
ΙΟ
ο
ο I 1 I I 1 1 III 1 1 ι 11 m l ι 1 1 1 1 III
ο
10 - 5 10- 10 -3 10-
Δι,,Ι 2
Figure 5: Cyclic stress strain curves according to multi level tests of the steel
StE-70.
StE 70 X0 CrNi 13 11
As / MP« 619 240
A l / Λ/Ρα 1148 2940
ns 0.027 0.056
nL 0.124 0.443
ill
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
Α € ρ / 2 =
( κ ) (6)
with ci =0.25 and C2 = 160 MPa.
The dashed line in figure 6 shows the data obtained in multi-level tests with in-
creasing amplitude (figure 5). There is good agreement at large amplitudes only.
ο
ο
rs
112
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
Since the incremental step test data depend on the maximum amplitude, there is
no unique relation between stress and strain. If, however, the shape of individual
stabilized hysteresis loops are compared with the CSSC curves from the incre-
mental step tests, there is a good agreement as required by the Masing behaviour.
Because of this the stress-strain relations obtained from incremental step tests
were used for life-time predictions only.
Many studies have shown that crack initiation occurs at special sites: inclusions,
grain boundaries, surface imperfections or slip bands with intrusions and extru-
sions. In the ferritic steel crack initiation at inclusions is dominant in biaxial tests.
Heitmann [11] reach the same conclusion for uni-axial loading of the same ma-
terial. For the stainless steel Ebi [16] and Tönnesen [17] have shown that at me-
dium amplitudes slip band crack initiation is dominant. Near the fatigue limit slip
bands are observed only in the intimate neigbourhood of twin boundaries. This re-
sult was explained by elastic imcompatibility stresses which occur at grain bound-
aries and also at twin boundaries due to elastic anisotropy (Neumann [18]). This
model explains also the striking feature that in a stack of twin lamella only every
other twin boundary is prone to cracking. These effects were also observed under
biaxial loading. Figure 7 gives an example.
113
Vol. 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
It is well known that after a certain distance of crack propagation along the slip
plane (stage I of crack growth) the crack plane switches to the plane of maximum
tensile stress (stage II of crack growth). In push-pull experiments all planes which
form an angle of 45° against the tensile direction (figure 8) are maximum shear
planes. Therefore, the change from stage I to stage II will, most likely, be detected
only by inspection of the fracture surface in the interior of the specimen. In torsion
experiments, however, there are only two sets of maximum shear planes and maxi-
mum tensile stress planes with different traces on the specimen surface such that
the transition from stage I to stage II is readily observed on the surface. Figure 9
shows a typical example in a torsion experiment in the low amplitude regime:
When the crack, which starts on the horizontal maximum shear stress plane, re-
aches a certain length, the growth mode switches to stage II. At the point of transi-
tion almost all cracks do branch onto the two possible planes of maximum tensile
stress. This behaviour can be explained as follows: At the critical length of the
stage I crack a transition to both possible maximum tensile stress planes is equally
likely. Under usual conditions branching would be suppressed because of the fol-
lowing: If the crack on one plane is, by chance, a little longer than the other, it will
Tension
Torsion
114
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
automatically reduce the stress concentration at the shorter crack, thus it will grow
faster than the shorter crack, and will in this instable way soon outrun the shorter
crack. This mechanism does not work in the torsion case. Both cracks will always
grow when the tensile stress is maximum across their crack plane. This occurs,
however, in the two different half cycles of the loading cycle. When one crack is
opened in tension, the other will be closed by the compressive normal stress on its
crack plane. In this sense both crack branches can grow independently in the two
different parts of the loading cycles.
The appearance of this kind of branching pattern critically depends on the number
of crack nuclei formed. In high amplitude loading (range I in figure 10) a large
number of crack nuclei is formed at the surface. Before the critical crack length for
stage II crack growth is reached, neighbouring cracks start to interact and there-
fore long cracks are formed by linkage of many small stage I cracks. In an inter-
mediate range II patterns resembling the typical crack branching configurations
occur (figure 10). However, these do not form by stage II crack growth but by link-
age of stage I cracks in an average direction of stage II cracks which gives the
cracks a stair-like course.
More or less the same behaviour was observed in the austenite. However, the
ranges I and II of figure 10 are shifted to larger strains. Below 0.5% range III-beha-
Figure 9: Crack growth in StE-70 during torsion, range III (fig. 10).
Δγ ρ /2 = 7.5 * 10-5, N l m m = 149000.
115
Vol. 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
viour is found. Because of the larger grain size and the well defined slip planes in
the austenite, a pure stage II crack growth is disturbed considerably by the interac-
tion with slip bands and grain boundaries. Figures 11 and 12 give examples of
cracks in the austenite. A similar behaviour was described by Bannantine and
Socie [19] in the stainless steel AISI 316.
The growth of individual cracks can be obtained only in the high-cycle fatigue
regime because then the crack nuclei are far enough apart. Short cracks which
grow in stage I show a crack growth rate which is independent of the crack length.
When the transition to stage II occurs there is a drop in the crack growth rate fol-
lowed by a strong increase in growth rate which is slowed down when the crack
depth becomes comparable with the wall thickness. Figure 13 shows a typical be-
haviour.
i i § >< X
1
Figure 10: Schematic diagram of the shape of fatigue cracks during torsion in
StE-70.
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
10 mm
Specimen Axis I I
When a tensile stress is superimposed onto the torsion, the plane of maximum
shear is rotated and the principal stresses on both maximum normal stress planes
become different. Stage I crack growth is modified accordingly. The cracks grow on
117
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
Q) —
iH
υ
>1
υ
R = -1
Λ Ο
"3 —
οΟ
9>
Figure 13: Crack growth rate during torsion as a function of the surface crack
length Δγ ρ /2 = 3.46 * ΙΟ'5, Νχ m m = 46100.
the inclined planes of maximum shear stress. If the amplitude is small enough to
avoid crack linkage the transition to stage II growth occurs as expected: The
branching configuration becomes asymmetrical because of the difference in the
magnitude of the principal stresses. When the biaxiality is small enough the second
branching is completely suppressed. Figure 14 shows an example of such a crack at
various stages of development. Also the crack growth rate is similar to that shown
in figure 13.
118
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
Λ/.Υ. SB 0 ,84
For materials which show Masing behaviour the AJ-integral is a measure of the
driving force for crack growth under cyclic loading. In a previous study [11] it was
shown that crack growth in cyclic push-pull tests can be described by a power law
dependence on the effective AJ-integral.
da/dN = CjAJ:}}
119
VoL 3, Nos. 2-3.1991 Life /Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
ι
ο
Ξ I Ι ! II lll| 1 1 1 1 Mill 1 1 1lll±
1
x
Ε StE-70 Xx -
X
a) ο
1-1
X
-η X
υ X
>1 X
U X -
X
S <" X*
5 ' * *
X*
—
Ξ χ **x
X
X
R = - 1
-
ι
ο
120
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
Ρτ (9)
Δτ/2
was used to express the effective shear stress range as a function of the total shear
stress range At and the absolute value of pT. The quantitative relation which fits
best to the data if given by
(10)
Are}J = Δτ * 0.350 ( \pT\ + 0.055) 0191
In figure 16 the crack growth is plotted as a function of A J e f f for the ferritic steel as
determined from experiments with constant plastic strain amplitude. The tensile
data are taken from [11]. The push-pull, torque and biaxial data are coincident for
Δ J eff > 2 5 0 J /m2 and fall on a common line. These data refer to stage II growth
which is obviously characterized by a power law dependency on AJ. The values of
the constants of the power law are given in figure 16.
Ξ /L —
υ
ä
υ
- StE-70 —
ε I
a.
Ο
IIIIII 1 1 1
/ J?*
- o o / ο
c
-υ tension
I
d ο
Ό 1 o torsion
IIIIII 1 I 1
X
1I 1IIIIII
ICO
biaxial, A fp
>
II
Δ
rv
II
Ο biaxial, X(p
TT
I
ο 1 ι ι 11 n i l ι ι ι ι ι nil ι 1 1 1 1 III
Δ J,jj j J/m2
Figure 16: Crack growth rate as a function of the effective AJ integral for the steel
StE-70 at various strain biaxialities and loading amplitudes.
121
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
The fast growth of small cracks is usually contributed to the lack of crack closure at
short cracks [20]. Ebi [18] has shown that this is not sufficient to explain the fast
growth in the austenite. Obviously, the fast growth is also due to the favourable
conditions for crack growth at those places at which crack initiation takes place
due to microstructural inhomogenities.
The ferritic steel (figure 18) shows a more complicated behaviour. Again there are
differences between push-pull and torsion tests. Therefore it must be concluded
122
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
that for both steels the equivalent plastic strain range is not sufficient to account
for the life data.
StE-70
Μ
I
ο
Ν?"
I x tension
ο
Ο torsion
Δ biaxial, Af(, =
Ο biaxial, Af(i = 4
ID
I
ο I I I III
2
10 103 104 10° 106
mill
Since crack growth can be described reasonably well by the AJ-integral (cf. figure
16) the life data may be plotted against the corresponding damage parameter
ZD = AJeJ}/a (11)
where AJ e ff is given by equation (8). In figure 19 the life data from the experiments
with constant plastic strain amplitude are given. Obviously, the data from push-
pull, torsion and biaxial tests fall on a common line which can be described quanti-
tatively for the ferritic steel as
(12)
V
/ 7 xx-1.46 >
„ ίη
Ν ι mm = 4.77 * 10 J/m'
123
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
The scatter in figure 18 is about 30%. For N i m m > 10^ cycles the elastic contribu-
tions to Z]3 dominate. This explains why there are no deviations in the high-cycle
fatigue range in figure 19 in contrast to the Coffin/Manson-law in figure 18. It is
important to notice that no empirical adjustments had to be made to describe the
life data of the various experiments with one common line. This may be due to the
fact that the Z j ) is a physically meaningful parameter suitable to describe the driv-
ing force on a crack under cyclic loading.
4 Conclusions
1) The cyclic stress strain curves depend only mildly on biaxiality. Cyclic deforma-
tion behaviour can be approximated by the Masing behaviour if stabilized hys-
teresis loops are used.
oo
Ο
ζ J 1 1 IIIIII 1 11 IIIIII 1 1 1 MILL I M M
1.46 —
-
\ Ν Iriim — 4.77*1013(^)
— —
StE-70
Δ\ —
Q
CS)
ο
Ο — —
x tension A
Ο torsion
- Δ biaxial
_J
ΙΟ 2 ΙΟ 3 ΙΟ4 ΙΟ 5 10 e
Figure 19: Relation between life time and the damage parameter Zj) for the steel
StE-70.
124
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
2) Crack initiation occurs in the ferritic steel within the first 10 % of the life. Thus
crack growth determines the fatigue life. In the austenitic stainless steel crack
initiation at the fatigue limit occurs preferentially at twin boundaries also under
biaxial loading.
3) The fast growth of small cracks on shear stress planes is nearly independent of
crack length. Long cracks grow along the planes of maximum normal stress. At
the transition from small to long cracks a reduced crack growth is observed.
4) The usual expression for the AJ integral was extended to the biaxial case with
the help of equivalent stresses and strains. Beyond a critical crack length crack
growth rate was found to be proportional to a power of the AJ integral. If the
life is plotted against the damage parameter Z j } = AJ e ff/a, the data fall on a
universal line independent of biaxiality and amplitude. Therefore, life predic-
tions can be made for biaxial loading cases by referring exclusively to uniaxial
data.
5 References
[1] Brown M.W. and Miller K.J., Α Theorie for Fatigue Failure under Multiaxial
Stress-Strain Conditions, Proc. Instn mech. Engrs, Vol.187, pp.745-755, 1973
[2] Lohr R.D. and Ellison E.G., Biaxial High Strain Fatigue Testing of lCr-Mo-V
Steel, Fatigue of Engineering Materials and Structures, Vol.3, pp. 19-37, 1980
[3] Kandil F.A., Brown M.W. and Miller K.J., Biaxial Low-Cycle Fatigue Frac-
ture or 316 stainless steel at Elevated Temperatures, The Metals Society,
Vol.280, pp.203-210, London, 1982
[4] Socie D.F., Waill L.A. and Dittmer D.F., Biaxial Fatigue of Inconel 718 In-
cluding Mean Stress Effects, Multiaxial Fatigue, ASTM 853, KJ.Miller and
M.W.Brown, Eds., American Society for Testing and Materials, Philadelphia,
pp.463-481, 1985
[5] Krempl E., Multiaxial Fatigue - Present and Future Methods of Correlation,
A.G.A.R.D. Conference Proceedings, No.155, pp. 5.1-5.12, August 1974
[6] Garud Y.S., A New Approach to the Evaluation of Fatigue under Multi-axial
Loadings, in Proceedings of the Symposium on Methods for Predicting Ma-
125
VoL 3, Nos. 2-3,1991 Life Prediction Under Biaxial Loading Based on the Growth Behaviour of Short Cracks
[7] Ellyin F., Cyclic Strain Energy Density as a Criterion for Multiaxial Fatigue
Failure, Biaxial and Multiaxial Fatigue, Eds. Brown M.W. and Miller K.J.,
E G F 3, Eng. Publications, London, pp.571-583, 1989
[8] Dowling N. and Begley J., ASTM STP 590, American Society for Testing and
Materials, pp.82-103, 1976
[10] Wiithrich C., The Extension of the J-Integral Concept to Fatigue, Int. Journ.
Fract., Vol.20, PP.35-37, 1982
[13] Heroux L.G. and Sullivan C.P., Trans.ASM, Vol.56, pp.861-863, 1963
[14] Nadai Α.,Theory of Flow and Fracture of Solids, 2nd edition, Vol.1, Ch.21
(McGraw-Hill Book Co., New York), 1950
[15] Neumann P. and Prochotta J., Shear Stresses in Thick Walled Tubular Speci-
mens under Torsion, to be published
[18] Neumann P. and Tönnessen Α., Cyclic Deformation and Crack Initiation,
Proc. Conf. Fatigue, Vol.1, pp.3-22, Charlotsville USA, 1987
[19] Bannantine J.A. and Socie D.F., Observations of Cracking Behavior in Ten-
sion and Torsion Low Cycle Fatigue, Low Cycle Fatigue, ASTM STP 942, pp.
J. Prochotta and P. Neumann Journal of the Mechanical Behavior of Materials
[20] Schubert R., Verformungsverhalten und Rißwachstum bei Low Cycle Fatigue,
Forschungsberichte VDI, Reihe 18: Mechanik/Bruchmechanik, Nr.73, 1989
127